Unmanned aerial vehicles (UAVs) have been widely used in industry and daily life, where safety is the primary consideration, resulting in their use in open outdoor environments, which are wider than complex indoor environments. However, the demand is growing for deploying UAVs indoors for specific tasks such as inspection, supervision, transportation, and management. To broaden indoor applications while ensuring safety, the quadrotor is notable for its motion flexibility, particularly in the vertical direction. In this study, we developed an improved probabilistic roadmap (PRM) planning method for safe indoor flights based on the assumption of a quadrotor model UAV. First, to represent and model a 3D environment, we generated a reduced-dimensional map using a point cloud projection method. Second, to deploy UAV indoor missions and ensure safety, we improved the PRM planning method and obtained a collision-free flight path for the UAV. Lastly, to optimize the overall mission, we performed postprocessing optimization on the path, avoiding redundant flights. We conducted experiments to validate the effectiveness and efficiency of the proposed method on both desktop and onboard PC, in terms of path-finding success rate, planning time, and path length. The results showed that our method ensures safe indoor UAV flights while significantly improving computational efficiency.

High-resolution remote sensing images contain rich color and texture information, but due to the inherent limitations of 2D data, achieving high-quality semantic segmentation remains a challenge. Multimodal data fusion technology has emerged as an effective approach to overcome this issue. To accurately capture the semantic information in remote sensing images, this study designs a multimodal fusion transformer-based DeepLabv3+ model for remote sensing semantic segmentation, named FTransDeepLab. Specifically, the network learns features from two modalities and is inspired by the DeepLab architecture. We extended the encoder by stacking the multi-scale Segformer, encoding the input images into highly representative spatial features. Additionally, we introduced the Multimodal Feature Rectification (MFR) module and the Multimodal Feature Fusion (MFF) module. The MFR, composed of a channel attention module and a spatial attention module, enhances the model’s ability to capture essential features and improves performance by focusing on both global and local contexts. The MFF module utilizes a cross-attention mechanism to optimize the feature fusion process, which enhances representation learning by facilitating the interaction between diverse information and integrates features from different modalities. Finally, in the decoding path, the extracted high-level features are concatenated with low-level features to optimize the feature representation and upsampled to restore the size of input image. Extensive results on two datasets, ISPRS Vaihingen and Potsdam, have confirmed that the proposed FTransDeepLab can achieve superior performance compared to state-of-the-art segmentation methods.